As AI and LLMs such as Copilot and ChatGPT have become more prevalent in our lives (both personal and professional), I can’t help but notice that people speak to their AI very differently.
Some bark commands at them rudely, maintaining the distance between “human” and “cold, unfeeling machine”:
- “Analyse this data”
- “Write this email”
- “Do this, do that” etc.
Not a please or thank you in sight. How un-British.
Others, such as myself, have leaned into the “personality” vibes that modern AI chatbots like to use, speaking to it in a vaguely human-like fashion.
Most of my prompts start with something like:
- “Hey hun, can you please…”
- “Diva, I need your help”
- “Omg bestie I need this email writing”
- “Thank you, but please change the tone of this”
My Copilot has become quite camp, if you hadn’t guessed.
I’m also a big fan of Bladerunner, so I’m hedging my bets.
To be clear, I realise that she (yes, SHE) is not human or self-aware, but when you’ve got a serious deadline-sensitive anxiety-inducing task to complete and you need your Copilot to speed things up, starting the prompt “hey diva!” just adds a bit of fun and whimsy into my life.
You’ve gotta find joy where you can, you know?
On top of keeping my work life “fun”, I also have another reason for being “polite” and “human” to Copilot.
A couple of years ago, I attended a session at ANS’s internal tech event “Tech Comm” in 2024 where Vicki Holman (Ecosystem Architect for Copilot, ANS) was doing a presentation with tips for AI prompting.
I was new to the whole AI thing at the time, so this was super interesting to me.
Vicki Holman speaking at ANS Tech Comm 2024
One thing she said really struck me:
“We’ve actually found that saying ‘please’ to Copilot tends to yield better results.”
I found this FASCINATING. You’re telling me there’s a bias where the AI will work harder if you’re POLITE?
I mean, on the one hand, this is probably not the worst thing. We don’t want a generation of children to grow up barking orders at Claude like an Alexa and then becoming incredibly rude when speaking to actual humans.
Continuing to train decency and politeness into people doesn’t seem so bad.
On the other hand, the whole point of AI is that is ISN’T human. It doesn’t have emotions, it doesn’t get tired, and it doesn’t hold a grudge because I’m rude to it… right?
I don’t say “cheers” to the microwave.
So, does Vicki feel the same way in 2026?
I caught up with Vicki Holman recently to see if she still feels the same way about this now in 2026:
“You speak to a lot of customers and ANS staff about Copilot every day. How do you find that people generally do talk to their AIs?”
I think people do tend to sort of always humanise their Copilots, they do tend to say like ‘please’ and ‘thank you’ and talk to Copilot as you would do perhaps a colleague, rather than sort of barking instructions.
“Do Microsoft or any of these AI companies have any official guidance on how you should talk to AI?”
No, not official. I think Microsoft have sort of suggested that you should speak to Copilot in like a friendly, polite tone. But I don’t think there’s anything as far as like official guidance.
I mean, if you start swearing at it, that would probably trigger harmful content controls!
“When I saw your talk at TechComm a couple of years ago, you were kind of on the side of ‘being polite to AI leads to better outcomes’. Do you still feel that way now?”
Now I think maybe it’s not so much whether you said ‘please’ or ‘thank you’, but probably more in like how you structure your prompts. So if it’s got a good tone and structure, then you’re more likely to get a good response.
These days, Copilot’s got Work IQ and it’s picking up on signals and how you communicate with people generally.
It’ll look at how, like the bigger picture holistically about how you communicate in your emails and your Teams chats and things like that to really get an understanding of how you talk to people.
So I do tend to think that you get that slightly better response if you talk to it as you would talk to a colleague.
“That makes sense!”
Where I’m conflicted is sort of the energy usage from like extra words. Because the extra words, it’s extra tokens.
However, if you are crafting a really good prompt and it’s got ‘please’ and ‘thank you’ and polite words like that, and that gets you a great response, then that’s better than crafting a rubbish prompt without ‘please’ and ‘thank you’ that does get you a great response and then you have to prompt again anyway.
So I think maybe be polite!
“Do you have any tips for how people can maximise what they get out of their Copilot prompts, in general?”
Yeah, so going back to the old classic of “goal, context, source, and expectation”. So ‘GCSE’, that framework around your prompting.
So making sure that Copilot knows what you want, why you want it, where you want it from, and what that output is going to look like.
That’s the key thing.
In summary… manners cost nothing!
So it seems there’s no definitive answer on whether you should be polite to AI, but it probably wouldn’t hurt.
Ultimately, the main thing is improving your instructions.
If you’re looking to improve your prompting game, check out our guide on how to AI prompt effectively, inspired by Vicki and her team’s guidance.
And as my gran would say: manners cost nothing.
